Autopsy your winners: finding the products behind a sales jump
A sales jump can hide very different product results. Compare ASIN dollar contributions and session changes to find what deserves investigation.
An unusually good day gets my attention as fast as a bad one. The total catches my eye. It doesn’t tell me where the money came from.
Sort by the dollar change
I used to download two periods from Amazon, map the products in a spreadsheet, put sessions, units and revenue side by side, and sort by dollar difference.
Here is one of those comparisons, with the product names changed and the figures left alone. Sales went from $20,624 to $38,853, up $18,229 or 88.4%. Sessions went from 2,529 to 3,409. Units from 1,016 to 1,683.
| Product | Dollar change | Share of net gain |
|---|---|---|
| Cold brew concentrate, 3-pack | +$4,391 | 24.1% |
| House blend, 12-pack | +$2,048 | 11.2% |
| Dark roast, 12-pack | +$1,901 | 10.4% |
| House blend, 4-pack | +$1,328 | 7.3% |
| Remaining products combined | +$8,561 | 47.0% |
| Total | +$18,229 | 100% |
Cold brew was the biggest contributor and still explained only a quarter of the gain. Plenty left to find.
The arithmetic: subtract period-one sales from period-two sales for each product, then divide by the account's net change. That share is not the product's growth rate. When gains and losses offset each other, one product’s share can exceed 100%. If the account’s net change is zero, that share is undefined; use dollar changes instead.
Keep marketplace, metric and ASIN level consistent across both periods, include products that appear in only one of them, check missing rows before treating them as zero, and reconcile the differences to the same-scope total. Don’t add parent totals to their children.
Change the sort when the question changes
Sessions are usually my next sort, by the difference between periods rather than by today's biggest traffic.
| Product | P1 sessions | P2 sessions | Gain | Growth |
|---|---|---|---|---|
| Flavored variety, 3-pack | 0 | 464 | +464 | N/A |
| Cold brew concentrate, 3-pack | 0 | 274 | +274 | N/A |
| Caramel, 3-pack | 0 | 183 | +183 | N/A |
| Mocha, 3-pack | 0 | 146 | +146 | N/A |
| Hazelnut, 3-pack | 0 | 72 | +72 | N/A |
Zero starting sessions means growth is undefined, so those rows get read as raw gains. For a nonzero example, the house blend 12-pack went from 30 to 76 sessions, up 153.3%.
Four of the five biggest session gains were flavored products. Together they added 865 product sessions. Those are visits, not a count of new customers. That gave me a second group to investigate alongside cold brew.
| View | What stood out | Next question |
|---|---|---|
| Revenue | Cold brew added $4,391 | What drove the biggest dollar gain? |
| Sessions | Flavored variety added 464; cold brew added 274 | Where did the additional visits come from? |
| Product group | Four flavored products added 865 sessions | Was something working across the group? |
Put the comparison in context
My spreadsheet labeled products "New Sale" and "Discontinued." Those labels can reflect a seasonal change, a stockout, an inbound delay or a launch. Activity in only one period tells you something happened, not why.
What to check next:
- Had these products been doing well for a while?
- Did the gain hold on the surrounding days?
- Did it hold against matching weekdays?
Revenue growing faster than units leaves a price or mix question open. Our comparison-window guide covers choosing the next period. This product breakdown sits inside the broader performance review.
Autopsy your winners
If we ran out of stock and I already know why, I don't need to rediscover the cause. Same for an explained Buy Box loss or a promoted period compared with a quiet one. The issue may still need action, but I don’t need to start the diagnosis over.
An unexpected gain is different. Something worked. I want to know what, so we can do it again or test whether we can. Contribution tells you where the movement was. Finding the cause takes the next check, including whether a new product added demand or drew it from a sibling. Start with the products responsible, then follow the evidence to the next move.